Machine Learning in the Corporate Bond Market and Beyond: A New Classifier

Abstract

Trade signing algorithms that rely on quote data, tick data or both have been used extensively to assign a trade as either a buy or a sell. Leveraging the availability of a large panel of signed trade data in the corporate bond market, we explore machine learning methods to uncover a new trade signing model that improves upon standard trade classification methods in both the bond and equity markets. We show that both the trading and information environment at the time of the trade critically affect the accuracy of existing trade classification rules in general, and also illustrate the importance of optimizing the feature set in correctly classifying trade direction. Importantly, our approach and the Random Forest algorithm we propose can be used in markets both with and without pre-trade transparency.

Keywords: Machine Learning, Trade Direction Classifiers, Trade Signing, Corporate Bonds, Equity Market, Big Data

JEL Classification: G0